Supporting data for "Deep Learning Functional Relationship Between Electron Density and Exchange-Correlation Potential"

The LOG files contain the data used to plot figures and tables in the thesis. The table LOG files contain the detailed normalize square difference (I values) between the calculated electron density and CCSD benchmark for different molecular systems. The other LOG files contains ZMP results (dipole moment for HF molecule and the potential difference between ZMP and B3LYP exchange-correlation potential), OEP genrated results (I values, density difference along dissociated bond/in XZ-plane, exchange-correlation potential comparsion between OEP and B3LYP and molecular properties (atomic forces, dipole moment and Mulliken charges)), the neural network training results (training loss, validation loss, root mean square error, and comparison between predicted and target potentials), FCNN results for different systems respectively (I values, density difference along dissociated bond/in XZ-plane,exchange-correlation potential comparsion between FCNN and B3LYP, and molecular properties (atomic forces, dipole moment and Mulliken charges)). Three softwares (Matplotlib, PySCF and Kspies) are used in this work. Matplotlib scripts used to reproduce the graphs are also included.

该数据集包含用于论文中图表和表格绘制的原始数据。表型LOG文件记录了不同分子体系中计算得到的电子密度与CCSD基准之间的详细归一化平方差（I值）。其他LOG文件则包含了ZMP结果（HF分子的偶极矩及ZMP与B3LYP交换关联势之间的电势差）、OEP生成结果（I值、沿解离键/在XZ平面上的密度差异、OEP与B3LYP之间的交换关联势比较以及分子性质（原子力、偶极矩和Mulliken电荷）），神经网络训练结果（训练损失、验证损失、均方根误差以及预测电位与目标电位之间的比较），以及针对不同系统的全连接神经网络（FCNN）结果（I值、沿解离键/在XZ平面上的密度差异、FCNN与B3LYP之间的交换关联势比较以及分子性质（原子力、偶极矩和Mulliken电荷））。本工作中使用了三种软件（Matplotlib、PySCF和Kspies）。此外，还包括了用于重现图表的Matplotlib脚本。